03 · Decision Intelligence
A decision-intelligence engine that tells you what to focus on now, and shows its reasoning. Not another task manager: a deterministic rule engine, a transparent scoring engine, and an aggregator that turns a messy backlog into a ranked, auditable shortlist.
Every productivity tool can store a backlog. None of the ones I used could tell me, with a straight answer, what to work on next, and why. Either they sorted by due date alone (everything urgent feels equally urgent the day before it's due), or they buried "smart" sorting behind an LLM call that produced a different answer every time you asked.
I wanted a system that behaves like a colleague who's thought it through: ranks the work, and can explain the ranking in plain numbers, every time, the same way.
I designed and built the full stack: the FastAPI scoring service, the PostgreSQL schema for tasks and weights, and the React frontend that surfaces the ranked shortlist with its reasoning attached.
Octostra is broader than a ranked task list. The app ties the personal focus queue to projects, sprint planning, team workload, recommendations and performance tracking.
Focus
The core view turns a backlog into the next few decisions, exposing score components so users can audit why each item appears.
Sprints
Planning is connected to execution: active sprint scope, progress, risk and board state are visible in the same workspace.
Team
The team view makes capacity visible, so the engine can highlight overload and match work to the people best placed to move it.
Performance
Performance analytics show whether the recommendation engine is improving throughput or simply reshuffling the backlog.
Each stage is a separate, testable unit. The scoring weights are configuration, not code, so the ranking can be tuned per user without touching the engine.
Tasks come in with metadata: due date, declared importance, estimated effort. Stored in PostgreSQL via the FastAPI API.
Deterministic rules normalize each field into a 0–100 sub-score: urgency from days-until-due, impact from declared importance, effort from estimated cost.
Sub-scores are combined with configurable weights into a single 0–100 total score per task, with each component retained for explanation.
Sorts by total score, returns the top 3–5 as the day's shortlist, and attaches the per-component breakdown that justifies each rank.
This is a small, faithful re-implementation of Octostra's ranking logic: urgency (45%), impact (35%), effort (20%). Add a task, watch it find its place, and click any row to see exactly why it ranked where it did.